Original Source:  Aunalytics

“Introduction

If you’ve been reading our blog regularly, you have noticed that we mention decision trees as a modeling tool and have seen us use a few examples of them to illustrate our points. This month, we’ve decided to go more in depth on decision trees—below is a simplified, yet comprehensive, description of what they are, why we use them, how we build them, and why we love them. (Does that make us the tree-huggers of the digital age? Maybe!)

What is a Decision Tree?

A decision tree is a popular method of creating and visualizing predictive models and algorithms. You may be most familiar with decision trees in the context of flow charts. Starting at the top, you answer questions, which lead you to subsequent questions. Eventually, you arrive at the terminus which provides your answer. (If you are unfamiliar with flow charts, this humorous image illustrates the concept.)

Decision trees tend to be the method of choice for predictive modeling because they are relatively easy to understand and are also very effective. The basic goal of a decision tree is to split a population of data into smaller segments. There are two stages to prediction. The first stage is training the model—this is where the tree is built, tested, and optimized by using an existing collection of data. In the second stage, you actually use the model to predict an unknown outcome. We’ll explain this more in-depth later in this post.

It is important to note that there are different kinds of decision trees, depending on what you are trying to predict. A regression tree is used to predict continuous quantitative data. For example, to predict a person’s income requires a regression tree since the data you are trying to predict falls along a continuum. For qualitative data, you would use a classification tree. An example would be a tree that predicts a person’s medical diagnosis based on various symptoms; there are a finite number of target values or categories. It would be tempting to simply conclude that if the information you are trying to predict is a number, it is always a regression tree, but this is not necessarily the case. Zip code is a good example. Despite being a number, this is actually a qualitative measure because zip codes are not calculated; they represent categories…”

https://www.aunalytics.com/2015/01/30/decision-trees-an-overview/

Contributors: Open Source
Categories: Industry Articles, SIG U Resource
SRC Type: Benchmarking and Analysis, Category Specific, Finance, Procurement Operations, Real Estate